基于神经网络的安全评价指标重要度判定方法及应用

安全评价指标重要度判定对合理进行安全管理和有效采取安全对策具有重要意义,它是一个多指标非线性分类问题,很难用数学公式进行描述。以往的判定方法由于受人为因素及模糊随机性影响,准确性较低。神经网络作为一种新技术,具有非线性分类、人工智能的特点。基于此,提出了一种基于人工神经网络的安全评价指标重要度判定方法。该方法最大特点是直接从学习后的网络连接权重中提取评价指标重要度信息。讨论了网络的拓扑结构,以及如何从学习后的网络权重中提取评价指标重要度信息的方法。应用数理统计方法消除了网络学习初始权重对评价结果的影响。用一实例对提出的方法进行了验证,分析了网络隐含层节点数对判定结果的影响。实验表明,该方法具有很强的操作性和较高的准确性。

Determination of Index Weight in Safety Assessment Based on Neural Network and Its Application

Accurate determination of index weight in safety assessment is of great importance in appropriate safety management and taking effective safety countermeasures. It is such a nonlinear categorization problem of multiple indexes difficult to be described with mathematical formulas. Affected by man-made factors and fuzzy randomness, the former methods for weight determination are levy in veracity. As a new technique, neural network has the characteristics of nonlinear categorization and artificial intelligence. It is on this basis that a new method based on artificial neural network is put forward. The utmost characteristic of the method is to pick up the infonmation of index weight of assessment directly from the network link weight after training. The network topological structure and the method for picking up information are discussed. The influence of primary network weight on final assessment result could be eliminated by symbolic statistics. The method is verified and exemplified, and the influence of the number of links in the hidden layer of network on final result is analyzed. Experiment shows that the method is highly operational with high accuracy.